Two distinct overstretched DNA structures revealed by ... · naked DNA. One intrinsic property is the thermodynamics of the transitions. ΔS and ΔH during DNA melting have been studied

Two distinct overstretched DNA structures revealedby single-molecule thermodynamics measurementsXinghua Zhanga,b,c, Hu Chenb, Hongxia Fub, Patrick S. Doylea,d,1, and Jie Yana,b,c,e,1

aBioSystems and Micromechanics (BioSyM), Singapore-MIT Alliance for Research and Technology, National University of Singapore, Singapore 117543;bMechanobiology Institute, National University of Singapore, Singapore 117411; cDepartment of Physics, National University of Singapore, Singapore117542; dDepartment of Chemical Engineering, Massachusetts Institute of Technology (MIT), 77 Massachusetts Avenue, Cambridge, MA 02139; andeCentre for Bioimaging Sciences, National University of Singapore, Singapore 117546

Edited by* Stephen C. Kowalczykowski, University of California, Davis, CA, and approved March 20, 2012 (received for review June 21, 2011)

Double-stranded DNA is a dynamic molecule whose structure canchange depending on conditions. While there is consensus in theliterature about many structures DNA can have, the state of highly-stretched DNA is still not clear. Several groups have shown thatDNA in the torsion-unconstrained B-form undergoes an “over-stretching” transition at a stretching force of around 65 pN, whichleads to approximately 1.7-fold elongation of the DNA contourlength. Recent experiments have revealed that two distinct struc-tural transitions are involved in the overstretching process: (i) ahysteretic “peeling” off one strand from its complementary strand,and (ii) a nonhysteretic transition that leads to an undeterminedDNA structure. We report the first simultaneous determination ofthe entropy (ΔS) and enthalpy changes (ΔH) pertaining to these re-spective transitions. For the hysteretic peeling transition, we deter-mined ΔS ∼ 20 cal∕ðK:molÞ and ΔH ∼ 7 kcal∕mol. In the case of thenonhysteretic transition, ΔS ∼ −3 cal∕ðK:molÞ and ΔH ∼ 1 kcal∕mol.Furthermore, the response of the transition force to salt concentra-tion implies that the two DNA strands are spatially separated afterthe hysteretic peeling transition. In contrast, the corresponding re-sponse after the nonhysteretic transition indicated that the strandsremained in close proximity. The selection between the two transi-tions depends on DNA base-pair stability, and it can be illustrated bya multidimensional phase diagram. Our results provide importantinsights into the thermodynamics of DNA overstretching and confor-mational structures of overstretched DNA that may play an impor-tant role in vivo.

entropy and enthalpy ∣ S-DNA ∣ ssDNA ∣ B-to-S transition

DNA can exist as a single-stranded polymer or a double-stranded helical structures. In cells, DNA primarily exists in

the stable B-form (B-DNA), which contains two strands that areassociated by Watson-Crick base-pairing interactions, and arestabilized by stacking interaction between adjacent base pairs.The transition from B-DNA to single-stranded DNA (ssDNA)is called DNA melting, and it is necessary for many fundamentalprocesses such as DNA replication, gene transcription, and DNAdamage repair. In vivo, DNA melting can occur with assistancefrom DNA helicases or ssDNA binding proteins (1, 2). In vitro,DNA melting can occur by directly heating or pulling the twocomplementary strands apart in a single-molecule unzipping ex-periment (3).

Double-stranded DNA can exist in several different structuresfrom the B-form, such as A-DNA and Z-DNA. These alternativestructures can be promoted under certain conditions (4, 5).dsDNA can also exist in elongated forms in the presence of DNAdamage repair proteins, such as RecA and Rad51 (6), or DNAintercalating ligands, such as the dyes YOYO-1 and ethidium bro-mide (7). Mechanical stretching of DNA may produce a similartransition.

A structural transition, referred to as the DNA overstretchingtransition, occurs at a force of around 65 pN. After this transition,DNA is stretched to about 1.7 times the contour length pertainingto the B-form (8, 9). Since its discovery in 1996, there has been a

debate about the mechanism of this transition and the natureof overstretched DNA. The central question is whether over-stretched DNA is (i) ssDNA due to force-induced melting of theduplex, or (ii) a unique elongated form of dsDNA (S-DNA) re-sulting from a hypothetical B-to-S transition (8, 9).

Both models have strengths and weaknesses in the interpreta-tion of experimental data. A series of experiments support force-induced melting that leads to one ssDNA strand under tensionthrough peeling from nicks or open ends of DNA or two sepa-rated single strands under tension through melting inside theDNA (internal melting) (10–15). Particularly, studies of the de-pendence of the transition force on temperature FovðTÞ have de-termined ΔS and ΔH during the transition in dye-free conditions(14). The values are in good agreement with the thermal meltingtransition (16), and they disfavor a nonmelting mechanism.Whether the overstretched DNA has only one strand or twostrands under tension can be studied by the dependence of thetransition force on the ionic strength (15). Two such experimentshave been reported. One study supports one strand (15), and theother study supports two strands under tension (10). Thus, basedon these experiments, peeling and melting have been proposed toexplain the DNA overstretching transition. Furthermore, force-induced DNA melting was also reported in full-atom moleculardynamics simulations (17).

In contrast, observations in a different series of experimentsimply a nonmelting mechanism. In Experiment 1, the force-response of overstretched DNA is inconsistent with that of onessDNA strand or two noninteracting ssDNA strands (18, 19). InExperiment 2, a second transition at an even higher force hasfrequently been observed that leads to final strand separationafter the 65-pN overstretching transition (19–22). The existenceof this second transition, which is definitely a melting transition,supports the notion that the first transition (at approximately65 pN) is not a melting transition (19–21). In Experiment 3, Paik,et al., and some of us, showed that end-blocked, torsion-uncon-strained DNA (which prevents peeling) still undergoes a nonhys-teretic DNA overstretching transition at approximately 65 pN(19, 23). A DNA melting mechanism, however, may also explainthese experimental results. For example, the unique force-response in Experiment 1 may represent an internally meltedDNA whose two strands are interacting with each other. The sec-ondary transition in Experiment 2 may be explained by breakingthe last base pairs holding the strands together due to the hetero-geneity in the DNA sequence (24). The nonhysteretic transition

Author contributions: J.Y. designed research; X.Z., H.F., and J.Y. performed research; X.Z.,H.C., P.S.D., and J.Y. analyzed data; and X.Z., P.S.D., and J.Y. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed. E-mail: [email protected] or [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1109824109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1109824109 PNAS ∣ May 22, 2012 ∣ vol. 109 ∣ no. 21 ∣ 8103–8108

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PHYS

ICS

Dow

nloa

ded

by g

uest

on

Janu

ary

16, 2

021

in Experiment 3 on the end-blocked, torsion-unconstrained DNAmay be an internal DNA melting transition. In addition, simula-tions and theoretical modeling studies suggest the existence ofnonmelted elongated dsDNA (25, 26).

Theoretically, these two conflicting mechanisms can be recon-ciled by the existence of two modes of DNA overstretching tran-sitions at approximately the same force and elongation (18),which may be the origin of confusion in the field. Consistent withthis view, a series of experiments by our lab have revealed a hys-teretic transition and a nonhysteretic transition at approximately65 pN, which can be selected or coexist via small changes infactors that affect DNA base pair stability (20). The hysteretictransition has been shown to be a peeling transition. Whetherthe nonhysteretic transition leads to a previously proposed non-melted “S-DNA” (8, 9, 18) or an internally melted DNA (10, 15)remains unclear. Due to this uncertainty, hereafter we refer theDNA after the nonhysteretic transition as “nonhysteretic over-stretched DNA.”

This research aims to provide new insights to the understand-ing of the following two major questions about DNA overstretch-ing transitions: (i) whether “nonhysteretic overstretched DNA” isinternally melted DNA, and (ii) how the selection between thehysteretic peeling transition and the nonhysteretic overstretchingtransition depends on experimental conditions. Crucial to thesuccess of this research, an unambiguous experimental indicatoris needed to judge whether the transitions are related to DNAmelting. One possible approach is to stain the overstretchedDNA with fluorescence dyes specific to ssDNA or dsDNA in or-der to visualize the DNA structural compositions directly (12).This approach, however, has a disadvantage of perturbing the sta-bility of DNA structures, and it may influence the experimentaloutcomes. Therefore, to eliminate possible effects of DNA bind-ing agents, it is important to study overstretching based on theintrinsic properties of the transitions and resulting structures of

naked DNA. One intrinsic property is the thermodynamics of thetransitions. ΔS and ΔH during DNA melting have been studiedextensively and are well characterized (16). If the nonhystereticoverstretching transition is not a DNA melting transition, thesevalues are expected to differ from the values pertaining to DNAmelting. In addition, intrinsic structural properties of an over-stretched DNA, such as the number (one or two) of strands undertension. In the case of two strands, their spatial proximity mayprovide further important insights. In this contribution, these in-trinsic properties are carefully examined for the hysteretic peelingtransition and the nonhysteretic overstretching transition withoutusing any DNA binding dyes.

As pointed out by Rouzina, et al. (15, 27), ΔS and ΔH duringDNA overstretching transition can be directly determined bymeasurements of FovðTÞ using the following equations: ΔS ¼−ð∂Fov∕∂TÞΔb and ΔG ¼ ΔΦþ ΔH − TΔS ¼ 0. Here Δb isthe DNA extension change per base pair during the transition(SI Appendix, Extension changes during transition), ΔΦ is the forcedependent free energy change that can be calculated with force-responses of B-DNA and overstretched DNA (SI Appendix,Entropy and enthalpy changes). According to recent studies fromour lab, there exist two distinct transitions based on whether hys-teresis exists (19, 20). Previous measurements of ΔS and ΔH,however, do not demonstrate any apparent distinct values (14).One possibility is that the hysteretic and nonhysteretic transitionsare DNA melting transition giving similar values of ΔS and ΔH.An alternative possibility is that the hysteretic and the nonhys-teretic transitions have distinct values of ΔS and ΔH. Underthose experimental conditions, however, DNA only underwentthe peeling transition; thus, the other transition type was notobserved. To test these possibilities, we remeasured FovðTÞ themover a wider temperature range and determined ΔS and ΔHduring respective transitions.

Fig. 1. Determination of Fov. (A) Schematic diagram of the transverse magnetic tweezers. A peltier chip was used to control the temperature. See SI Appendix,Magnetic tweezers measurements and SI Appendix, Temperature control and measurement for details. (B) DNA overstretching transition in the nonhysteretictransition represented by force-extension (top) and force-variance (bottom) curves measured at 15 °C in 500 mM NaCl (pH 7.5). The threshold value of varianceapproximately 500 nm2 (red line) in the force-variance curve was used to determine the onset of the transition (SI Appendix, Determination of the transitionforce). (C) DNA overstretching transition in the pure nonhysteretic transition (12–20 °C) and in the transition that contains hysteretic transition (22–24 °C). Insetshows the force-variance curves for four temperatures close to the onset of the transition, the red line corresponds to variance approximately 500 nm2.

8104 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1109824109 Zhang et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

16, 2

021

In addition to the measurements of FovðTÞ, important hintsto possible structures of an overstretched DNA can be obtainedby studying Fov as a function of ionic strength FovðI∕I0Þ (15).Here I is ionic strength, which is also the concentration of NaClin this research, and I0 ¼ 1 M, the standard ionic strength. FromFovðI∕I0Þ, a linear relation as a function of lnðI∕I0Þ exists with aslope ∂Fov∕∂ lnðI∕I0Þ ¼ νðkBT∕lBÞ for I ≪ I0. lB ∼ 0.71 nm isthe Bjerrum length in water at room temperature. The structuralcoefficient ν is approximately 1.2 if the transition leads to onestrand under tension while the other recoils (i.e., peeling), andν is approximately 0.5 if the two strands are tightly associated withan interstrand distance considerably less than the Debye screen-ing length. In this research, we also remeasured FovðI∕I0Þ to seewhether there exist distinct values of ν during the respective twotransitions. Finally, phase diagrams for the selection of transitionsare constructed from these results.

ResultsOur results were based on measurements of FovðTÞ andFovðI∕I0Þ using a transverse magnetic tweezers setup (28)(Fig. 1A, SI Appendix, Magnetic tweezers measurements, and SIAppendix, Temperature control and measurement). In our experi-ments, Fov is determined at the onset of the transition (Fig. 1 Band C) in order to attribute the force to a specific transition (SIAppendix, Determination of transition types). Analogous to otherphase transitions, a clear signature at the onset of the overstretch-ing transition is a dramatic increase in extension fluctuations.In our experiments, Fov is defined as the force where the varianceof the DNA extension increased to 500 nm2.

To find the onset transition, cycling between a force belowthe transition force and a series of increasing higher forces areperformed (SI Appendix, Fig. S4). At each of the higher forces,the DNA is held for 10 s (Figs. 1–4), during which the DNA ex-tension and variance are measured. The force-extension andforce-variance curves in Fig. 1 B and C were obtained by thismethod. If a variance of greater than 500 nm2 is found, the cor-responding force is identified as Fov. In addition to determiningFov, force cycling allows us to determine the transition types.In the peeling transition, hysteresis in extension change will beobserved due to the slow reannealing process that occurs at thelower forces; whereas, in the nonhysteretic transition, no hyster-esis in extension change will be observed due to much faster tran-sition kinetics (19, 20, 29).

Fig. 1C inset shows that the variance monotonically increasesas force increases in the nonhysteretic transition but not in thehysteretic peeling transition. This difference is caused by the slowstochastic nature of the peeling transition (19, 20, 29). Thus,determination of the transition force will have a larger variationin the hysteretic peeling transition than it will have in the non-hysteretic transition (SI Appendix, Fig. S6).

Using the above method, Fov was determined at differenttemperatures from which ΔS and ΔH could be calculated.Fig. 2A shows FovðTÞ measured in 150 mM NaCl and pH 7.5. Apiecewise linear temperature response was revealed with twodistinct slopes: ∂Fov∕∂T ∼ 0.12 pN∕K from 11 °C to 18 °C, wherethe nonhysteretic transition was determined, and ∂Fov∕∂T∼−0.77 pN∕K at T greater than 18 °C, where the hysteretic peelingtransition was determined. Switching from nonhysteretic transi-tion to hysteretic peeling transition as temperature increases isconsistent with an earlier observation that the level of hysteresiscan be suppressed by lowering temperature (30). We emphasizethat the approximately 18 °C switching temperature (temperatureat which the transition switches from the nonhysteretic transitionto the hysteretic peeling transition) observed here is likely a re-sponse of the less stable AT-rich DNA region (SI Appendix,Determination of the transition force).

Fig. 2B shows another two independent experiments per-formed in 10 mM (blue) and 500 mM (red) NaCl. In 10 mM

Fig. 2. Measurements of FovðTÞ including error bars in both force (SIAppendix, Fig. S6) and temperature (SI Appendix, Fig. S2D). In the nonhysteretictransition, Fov is denoted by colored filled squares and fitted to a linear function(solid line). In the hysteretic transition, Fov is denoted in colored open squares,and it is fitted to a linear function (dashed line). (A) FovðTÞ in 150 mMNaCl andpH 7.5 (black). A linear relation with a slope of approximately 0.12 pN∕K wasdetermined in the nonhysteretic transition (T < 18 °C). The slope remainsunchanged if the midpoint or the terminus of the transition was used to de-termine the transition force (SI Appendix, Determination of the transitionforce). A different linear relation with a slope of −0.77 pN∕K was determinedin the hysteretic transition (T > 18 °C). The hysteretic transition has beenagreed to be peeling of one strand from the other, while the nonhysteretictransition leads to an undetermined DNA structure, which are illustrated inthe panel. (B) In 500 mM NaCl (red), a similar piecewise linear relation was ob-tained with a slope of approximately 0.10 pN∕K in the nonhysteretic transitionand approximately −0.44 pN∕K in the hysteretic transition. In 10 mM NaCl(blue), only the hysteretic transition occurred and a single linear region wasobserved with a slope of approximately −0.92 pN∕K. The data obtained in150 mM (Fig. 2A) are also plotted for comparison (black). More independentexperiments obtained from other DNA molecules are shown in SI Appendix,Fig. S8. (C) The effects of Mg2þ to DNA overstretching in 150 mM NaCl andpH 7.5. Without Mg2þ (dark cyan), switching from the nonhysteretic transitionwith a slope of approximately 0.08 pN∕K to the hysteretic transition with aslope of approximately −0.68 pN∕K occurred at approximately 18.5 °C. In thepresence of 5mMMg2þ (magenta), switching from the nonhysteretic transitionwith a slope of approximately 0.08 pN∕K to the hysteretic transition with aslope of approximately −0.62 pN∕K occurred at approximately 19.4 °C.

Zhang et al. PNAS ∣ May 22, 2012 ∣ vol. 109 ∣ no. 21 ∣ 8105

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PHYS

ICS

Dow

nloa

ded

by g

uest

on

Janu

ary

16, 2

021

NaCl, the transition was entirely the hysteretic peeling transitionin the experimental temperature range with a slope of approxi-mately −0.92 pN∕K; whereas, in 500 mM NaCl, a piecewise lin-ear temperature response similar to that in Fig. 2A was observedwith a slope of approximately 0.10 pN∕K in the nonhysteretictransition and approximately −0.44 pN∕K in the hysteretic peel-ing transition. To better extrapolate these results to in vivo con-ditions, where magnesium exists in a mM concentration range, wealso studied the effects of magnesium. In 150 mM NaCl, Fig. 2Cshows that the piecewise linear temperature response still existsin 5 mM MgCl2, with similar slopes to the experimental dataobtained in the absence of magnesium for the same DNA. Theapparent effects of magnesium are that: it increases FovðTÞ byapproximately 1.5 pN, which is in agreement with previousstudies (31), and it increases the switching temperature by lessthan 1 °C.

Multiple independent experiments (SI Appendix, Fig. S8)have yielded the average and standard deviation of the slopesof 0.10� 0.02 pN∕K in 500 mM NaCl (ten experiments) and0.12� 0.02 pN∕K in 150 mM NaCl (six experiments) in the non-hysteretic transition; approximately −0.45� 0.05 pN∕K in500 mM NaCl (three experiments), and approximately −0.67�0.11 pN∕K in 150 mM NaCl (five experiments) in the hystereticpeeling transition. These slopes allowed us to calculate ΔS andΔH per base pair during the nonhysteretic transition and thehysteretic peeling transition (SI Appendix, Entropy and enthalpychanges) and compare them with that determined in the thermalmelting transition (averaged over a sequence with 50% GC con-tent) (16) shown in Table 1.

It is of interest to know whether the two strands of DNA are inclose proximity to each other after the two respective transitions.

Therefore, we studied FovðI∕I0Þ. Fig. 3 shows two independentexperiments at 11 °C (black) and 23 °C (red). At 11 °C, Fov wasfound to be a piecewise linear function of lnðI∕I0Þ, with twodistinct slopes: 2.9� 0.1 pN in greater than 20 mM NaCl, wherethe nonhysteretic transition was determined, and 5.7� 0.1 pNin less than 20 mM NaCl where the hysteretic peeling transitionwas determined. These slopes correspond to ν ¼ 0.53� 0.02 inthe former and ν ¼ 1.03� 0.02 in the latter. According to thepredictions by Rouzina, et al. (15), ν ¼ 0.53� 0.02 infers thatthe interdistance between the strands of the overstretched DNAis less than one Debye length (approximately 1 nm at 100 mMNaCl). This result suggests that the two strands of the “non-hysteretic overstretched DNA” are likely in close proximity.Moreover, ν ¼ 1.03� 0.02 observed in the hysteretic peelingtransition is close to the theoretically predicted value of 1.2 forthe hysteretic peeling transition (15). In another experiment at23 °C, the transition was determined to be the hysteretic peelingtransition for 1 mM < I < 500 mM. The corresponding slope is6.6� 0.16 pN and ν ¼ 1.15� 0.02. As shown in Fig. 3, the linearrange is only up to 100 mM; therefore, our fittings are up to100 mM NaCl-the same range as that used by Wenner, et al. (10).

Based on the experimentally determined force responses ofthe respective DNA states (B-DNA, ssDNA, or “nonhystereticoverstretched DNA”) (SI Appendix, Extension changes duringtransition), ΔS and ΔH during the DNA melting transition ob-tained in previous DNA thermal melting transition (16), andΔS and ΔH during the nonhysteretic transition measured in thisresearch (Table 1), as well as FovðI∕I0Þmeasured in this research(Fig. 3), we can construct phase diagrams to predict the statesof a DNA molecule with open ends or nicks and the selectionof the transitions as a function of external force F, temperatureT and ionic strength I (details in SI Appendix, Phase diagrams).

For clarity, we first consider the phase diagram projectedonto the F-T plane for a fixed ionic strength of 150 mM. Theboundary between the B-DNA and ssDNA (solid colored lines)where the free energy change ΔGB-ssðF; TÞ ¼ 0 can be calcu-lated from existing entropy and enthalpy data obtained fromDNA thermal melting experiments (16). This boundary willvary with GC content. Because there is no existing free energydata of the nonhysteretic transition, and our prior studies haveshown that the nonhysteretic transition is insensitive to GC con-tent (19, 20), we used the entropy and enthalpy changes in Table 1to calculate the boundary by ΔGB-NHOðF; TÞ ¼ 0, where NHOrefers to “nonhysteretic overstretched” for short. The boundarybetween “nonhysteretic overstretched DNA” and ssDNA(dashed line) is calculated by ΔGNHO-ssðF; TÞ ¼ ΔGB-ssðF; TÞ-ΔGB-NHOðF; TÞ ¼ 0. These three lines then determine the phaseboundaries of the system and meet at a triple point that corre-sponds to the switching temperature that was previously intro-duced. The data obtained from studies of FovðTÞ in Fig. 2 Aand C are replotted in Fig. 4A for comparison.

We have mentioned that the selection of the transitions de-pends on factors that affect DNA base pair stability. Analogousto the phase diagram, we can construct a phase diagram for theselection of the transitions. For a given GC content, the point atwhich the change from a nonhysteretic transition to a hystereticpeeling transition is given by the switching temperature or triplepoint. Using similar calculations as above, we can calculate a linein the I-T plane that divides these two transitions. Each line in

Table 1. Comparison of ΔS and ΔH between our results and that reported in thermal melting experiments

QuantitiesOur data, nonhysteretic

transition Our data, hysteretic transitionSanta Lucia (16), thermal

melting

I mM 150 500 10 150 500 10 150 500ΔS cal∕ðK:molÞ −3.8 ± 0.6 −3.0 ± 0.7 26.4 21.2 ± 3.5 14.2 ± 1.6 24.7 23.2 22.5ΔH kcal∕mol 0.9 ± 0.2 1.1 ± 0.2 8.6 7.7 ± 1.6 5.5 ± 1.0 8.2 8.2 8.2

Fig. 3. Measurements of FovðI∕I0Þ at two temperatures: 11 °C (open blacksquares for the hysteretic transition and filled black squares for the non-hysteretic transition) and 23 °C (red squares for the hysteretic transition).At 11 °C, a piecewise linear function of lnðI∕I0Þ was observed separated bya switching ionic strength of approximately 20 mM NaCl. It is characterizedby a smaller slope of 2.9� 0.1 pN (black solid line) in the nonhysteretic transi-tion and a larger slope of 5.7� 0.1 pN (black dashed line) in the hysteretictransition. At 23 °C, only the hysteretic transition was observed. A single lin-ear region with a slope of 6.6� 0.1 pN (red dashed line) was observed. Fromthese slopes ν was calculated. The lines shown in the figure are linear fits inthe respective transitions up to I ∼ 100 mM excluding the shadowed areawhere the theory is not applicable. Note: the standard deviation in forceis plotted as error bars in the figure, which are similar to the symbol sizeso they are not apparent in the figure.

8106 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1109824109 Zhang et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

16, 2

021

Fig. 4B corresponds to a different GC percentage, and it dividesthe I-T plane into the hysteretic peeling transition region (aboveof the line) and the nonhysteretic transition region (below theline). The triple points obtain from different experiments areplotted in Fig. 4B for comparison. The filled circles obtained onthe same DNA, while open circles are obtained from ten otherdifferent DNA molecules. Fig. 4 helps to emphasize that the ex-perimentally observed transition is sensitive to temperature, ionicstrength and GC content.

DiscussionWe have shown that FovðTÞ and FovðI∕I0Þ have distinct trends inthe nonhysteretic transition and the hysteretic peeling transition.ΔS and ΔH determined in the hysteretic peeling transition areconsistent with those measured in DNA thermal melting transi-tion experiments (16). The slight difference between our data andthose from DNA thermal melting could be explained by a finiteheat capacity change during DNAmelting (SI Appendix, Effects ofheat capacity change) (14, 15).

Striking results were found in the nonhysteretic transition.ΔS is a small negative value, which may suggest an ordered “non-hysteretic overstretched DNA” structure that has slightly lower

entropy than B-DNA together with surrounding water and iondistributions. The small ΔH value of approximately 1 kcal∕molis about one order of magnitude smaller than that measuredin thermal melting or hysteretic peeling transition. In addition,our study of FovðI∕I0Þ was consistent with a picture that thetwo strands in the nonhysteretic overstretched DNA are closeto each other at an interstrand distance within the Debye screen-ing length (15).

One important question remains regarding the exact structureof the nonhysteretic overstretched DNA. We considered twopossibilities: (i) the structure could be some new regular dou-ble-stranded structure with regular short-ranged bonds and resi-dual helicity (i.e., the previously proposed “S-DNA”), or (ii) thestructure could be the two separated strands with broken hydro-gen bonds. These two melted strands, however, can still interactwith each other strongly via electrostatic and steric interactions.We cannot draw a firm conclusion between these two possibilitiesbecause ΔS and ΔH during the force-induced DNA internalmelting transition were not directly measured.

Providing the final answer regarding the structure of the non-hysteretic overstretched DNA is not the purpose of this research.The main point of this research was to show that there exist twotransitions that have distinct entropy and enthalpy changes duringoverstretching of DNA with open ends or nicks; however, it isinteresting to note these results can be explained by the existenceof a nonmelting DNA overstretching transition, which warrantsfurther study. It is also worthwhile to mention a few previousexperiments that may be related to this research. It has beenfound that torsion-constrained DNA did not undergo over-stretching transition at approximately 65 pN unless the DNAis underwound (12). This result is also consistent with resultsobtained from another single-DNA stretching experiment by Bry-ant, et al. (32), and it is consistent with the high resolution atomicforce microscopy imaging of DNA overstretched by molecularcombing method (33).

These results raise interesting questions regarding the physio-logical relevance of the DNA overstretching transition. The hys-teretic peeling transition is sensitive to factors that affect DNAbase pair stability, and the transition force can be as low as 40 pNin 150 mMNaCl for AT-rich DNA at 37 °C (Fig. 4A). This force isclose to the force range that can be generated by a single RNApolymerase (34) or DNA polymerase (35) in the force range of20–40 pN. In comparison, the nonhysteretic transition is muchless sensitive to factors that affect base pair stability. Accordingto the predictions in Fig. 4B, the nonhysteretic transition mayoccur at greater than 25 °C for GC-rich DNA. The approximately60 pN transition force is about 30 pN greater than the reportedforce range that can be generated by RNA polymerase (34) orDNA polymerase (35). In the presence of DNA intercalators,however, it is known that elongation of double DNA requires lessforce. For example, recent experiments showed that the presenceof a YOYO-1 force of a few picoNewtons could elongate DNAcontour by approximately 1.5-fold (36). Although the structureof the nonhysteretic overstretched DNA remains unknown, weimagine that DNA bound with YOYO-1 may resemble the DNAstructure because it is only 10% shorter. In cells, DNA-distortingproteins play important roles in processing information in DNAand in organizing chromosome DNA. Among these proteins,many of them use side chain intercalation to distort the DNAbackbone (37). Therefore, binding of these proteins may alsobe susceptible to DNA tension.

Materials and MethodsRefer to SI Appendix for details of the DNA construct, magnetic tweezersmeasurements, temperature control and measurement, determination oftransition types, determination of the transition force, extension changesduring transition, entropy and enthalpy changes, phase diagrams, elimina-tion of thermal expansion effects, convection in the flow channel, and effectsof heat capacity change.

Fig. 4. Phase diagrams for states and transitions. (A) The phase diagramof B-DNA, ssDNA, and the nonhysteretic overstretched DNA (referred asNHO-DNA in the figure) projected onto the F-T plane in 150 mM NaCl. Foreach GC percentage, the solid line is the boundary between B-DNA and ssDNAand the dotted line is the boundary between the nonhysteretic overstretchedDNA and ssDNA. The point where the two boundaries meet is the triple point.The gray dashed line to the left of the triple point is the boundary betweenB-DNA and the nonhysteretic overstretched DNA. The data from Fig. 2A and Care plotted together in the same symbols for comparison. (B) The selectionof the transitions as a function of ionic strength, temperature, and sequence.Predicted phase boundaries for different GC contents are shown in differentcolors. In the region above a line, the transition will be via hysteretic peelingtransition, and the region below, a nonhysteretic transition will occur. Experi-mental data of the dependence of the switching temperature on ionicstrength from the same DNA molecule (filled circles) and from other tenDNA molecules (open circles) are also plotted for comparison.

Zhang et al. PNAS ∣ May 22, 2012 ∣ vol. 109 ∣ no. 21 ∣ 8107

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PHYS

ICS

Dow

nloa

ded

by g

uest

on

Janu

ary

16, 2

021

ACKNOWLEDGMENTS. We are grateful to John Marko (Northwestern Univer-sity), Ioulia Rouzina (University of Minnesota), and Stephen Kowalczykowski(UC Davis) for stimulating discussions. We also thank Michael Sheetz(Columbia University), Ioulia Rouzina (University of Minnesota), MichelleWang (Cornell University), Johan van der Maarel, and Ci Ji Lim (National

University of Singapore) for proofreading our manuscript. This work wassupported by the Ministry of Education of Singapore under Grant MOE2008-T2-1-096 by the Mechanobiology Institute at National University of Singaporeand by Singapore-MIT Alliance for Research and Technology at NationalUniversity of Singapore.

1. Caruthers JM, McKay DB (2002) Helicase structure and mechanism. Curr Opin StructBiol 12:123–133.

2. Iftode C, Daniely Y, Borowiec JA (1999) Replication protein A (RPA): the eukaryotic SSB.Crit Rev Biochem Mol 34:141–180.

3. Essevaz-Roulet B, Bockelmann U, Heslot F (1997) Mechanical separation of the com-plementary strands of DNA. Proc Natl Acad Sci USA 94:11935–11940.

4. Franklin RE, Gosling RG (1953) Molecular configuration in sodium thymonucleate.Nature 171:740–741.

5. Wang AHJ, et al. (1979) Molecular-structure of a left-handed double helical DNAfragment at atomic resolution. Nature 282:680–686.

6. Kowalczykowski SC, Eggleston AK (1994) Homologous pairing and DNA strand-exchange proteins. Annu Rev Biochem 63:991–1043.

7. Sischka A, et al. (2005) Molecular mechanisms and kinetics between DNA and DNAbinding ligands. Biophy J 88:404–411.

8. Smith SB, Cui Y, Bustamante C (1996) Overstretching B-DNA: the elastic response ofindividual double-stranded and single-stranded DNA molecules. Science 271:795–799.

9. Cluzel P, et al. (1996) DNA: an extensible molecule. Science 271:792–794.10. Wenner JR, Williams MC, Rouzina I, Bloomfield VA (2002) Salt dependence of the elas-

ticity and overstretching transition of single DNA molecules. Biophys J 82:3160–3169.11. Williams MC, Wenner JR, Rouzina I, Bloomfield VA (2001) Effect of pH on the over-

stretching transition of double-stranded DNA: evidence of force-induced DNA melt-ing. Biophys J 80:874–881.

12. van Mameren J, et al. (2009) Unraveling the structure of DNA during overstretchingby using multicolor, single-molecule fluorescence imaging. Proc Natl Acad Sci USA106:18231–18236.

13. Shokri L, McCauley MJ, Rouzina I, Williams MC (2008) DNA overstretching in thepresence of glyoxal: structural evidence of force-induced DNA melting. Biophys J95:1248–1255.

14. WilliamsMC,Wenner JR, Rouzina I, Bloomfield VA (2001) Entropy and heat capacity ofDNA melting from temperature dependence of single molecule stretching. Biophys J80:1932–1939.

15. Rouzina I, Bloomfield VA (2001) Force-induced melting of the DNA double helix. 2.Effect of solution conditions. Biophys J 80:894–900.

16. SantaLucia J, Jr (1998) A unified view of polymer, dumbbell, and oligonucleotide DNAnearest-neighbor thermodynamics. Proc Natl Acad Sci USA 95:1460–1465.

17. Santosh M, Maiti PK (2009) Force induced DNA melting. J Phys-Condens Mat21:034113.

18. Cocco S, Yan J, Leger JF, Chatenay D, Marko JF (2004) Overstretching and force-drivenstrand separation of double-helix DNA. Phys Rev E 70:011910.

19. Fu H, et al. (2011) Transition dynamics and selection of the distinct S-DNA and strandunpeeling modes of double helix overstretching. Nucleic Acids Res 39:3473–3481.

20. Fu H, Chen H, Marko JF, Yan J (2010) Two distinct overstretched DNA states. NucleicAcids Res 38:5594–5600.

21. Rief M, Clausen-Schaumann H, Gaub HE (1999) Sequence-dependent mechanics ofsingle DNA molecules. Nat Struct Biol 6:346–349.

22. Danilowicz C, et al. (2009) The structure of DNA overstretched from the 5′5′ endsdiffers from the structure of DNA overstretched from the 3′3′ ends. Proc Natl AcadSci USA 106:13196–13201.

23. Paik DH, Perkins TT (2011) Overstretching DNA at 65 pN does not require peeling fromfree ends or nicks. J Am Chem Soc 133:3219–3221.

24. Williams MC, Rouzina I, McCauley MJ (2009) Peeling back the mystery of DNA over-stretching. Proc Natl Acad Sci USA 106:18047–18048.

25. Lebrun A, Lavery R (1996) Modelling extreme stretching of DNA. Nucleic Acids Res24:2260–2267.

26. Haijun Z, Yang Z, Zhong-can O-Y (1999) Bending and base-stacking interactions indouble-stranded DNA. Phys Rev Lett 82:4560–4563.

27. Rouzina I, Bloomfield VA (2001) Force-induced melting of the DNA double helix 1.Thermodynamic analysis. Biophys J 80:882–893.

28. Yan J, Skoko D, Marko JF (2004) Near-field-magnetic-tweezer manipulation of singleDNA molecules. Phys Rev E 70:011905.

29. Gross P, et al. (2011) Quantifying how DNA stretches, melts and changes twist undertension. Nat Phys 7:731–736.

30. Mao H, Arias-Gonzalez JR, Smith SB, Tinoco I, Jr, Bustamante C (2005) Temperaturecontrol methods in a laser tweezers system. Biophys J 89:1308–1316.

31. Fu H, Chen H, Koh C, Lim C (2009) Effects of magnesium salt concentrations on B-DNAoverstretching transition. Eur Phys J E 29:45–49.

32. Bryant Z, et al. (2003) Structural transitions and elasticity from torque measurementson DNA. Nature 424:338–341.

33. Maaloum M, Beker A-F, Muller P (2011) Secondary structure of double-stranded DNAunder stretching: Elucidation of the stretched form. Phys Rev E 83:031903.

34. Davenport RJ,Wuite GJ, Landick R, Bustamante C (2000) Single-molecule study of tran-scriptional pausing and arrest by E. coli RNA polymerase. Science 287:2497–2500.

35. Wuite GJ, Smith SB, YoungM, Keller D, Bustamante C (2000) Single-molecule studies ofthe effect of template tension on T7 DNA polymerase activity. Nature 404:103–106.

36. Gunther K, Mertig M, Seidel R (2010) Mechanical and structural properties of YOYO-1complexed DNA. Nucleic Acids Res 38:6526–6532.

37. Werner MH, Gronenborn AM, Clore GM (1996) Intercalation DNA kinking, and thecontrol of transcription. Science 271:778–784.

8108 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1109824109 Zhang et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

16, 2

021